90 likes | 216 Views
Where to go from here?. Get real experience building systems! Opportunities: 496 projects More projects: http://netsyslab.ece.ubc.ca 491r/571r class next term Theme: Harnessing massively multicore systems (e.g., GPUs) http://www.ece.ubc.ca/~matei/EECE571/ USRA funding for summer
E N D
Where to go from here? Get real experience building systems! Opportunities: • 496 projects • More projects: http://netsyslab.ece.ubc.ca • 491r/571r class next term • Theme: Harnessing massively multicore systems • (e.g., GPUs) • http://www.ece.ubc.ca/~matei/EECE571/ • USRA funding for summer • (applications due January 2011)
A few projects … http://netsyslab.ece.ubc.ca • The top 1% of searchers performs a full 13% of all searches in a given month. If you extend this to the top 20% the number of queries increase to roughly 70%."Read more at:"http://glinden.blogspot.com/2007/11/who-cares-about-grandma.html"Although these relations are not news, the argument that follows and the link with advertisement revenues are really interesting.The goal of this project is to understand which queries are more likely to lead to advertising revenue using the search engine data Microsoft has (partially) made available [traces from their Live Search engine] Predict advertised item popularity?
Platform Example – Argonne Blue Gene/P 2.5K IO Nodes 160K cores GPFS IO rate : 8GBps = 51KBps / core !! 10 Gb/s Switch Complex Torus Network 24 servers 2.5 GBps per node 3D Torus 850 MBps per 64 nodes Tree The central storage is a bottleneck There are underutilized resources close to application
MosaStore Evaluation Zhang et. al., “Design and Evaluation of a Collective I/O Model for Loosely-coupled Petascale Programming”, MTAGS ’09. Overall: 1.52x
A few projects @ NetSysLab • P2P data storage system (MosaStore) • Application-level GPU harnessing • Online social systems
StoreGPU System design: balancing act in a multi-dimensional space. GPUs dramatically change the computation cost landscape. 10x FLOPS, 10x Memory bandwidth, yet same cost! Q: Does the 10x reduction in computation costs GPUs offer change the way we design/implement (distributed) storage system?
Data deduplication System -- Prototype Evaluation no-SD SD-CPU SD-GPU • 76% improvement in write throughput • No negative impact on concurrent applications Throughput (MB/s) Checkpointing a BLAST application 100 times [S. Al-Kiswany, A. Gharaibeh, S. Gopalakrishnan, and M. Ripeanu, “A GPU Accelerated Storage System”, Submitted to NSDI ‘10.]
Characterizing Online Social Systems • CiteULike, Flickr, YouTube, • Patterns of production/consumption of information are relatively unexplored • Usage patterns inform system design • Recommendation • Content pre-fetching • Spam detection
Where to go from here? Get real experience building systems! Opportunities: • 496 projects • More projects: http://netsyslab.ece.ubc.ca • 491r/571r class next term • Autonomic Systems • http://www.ece.ubc.ca/~matei/EECE571/ • USRA funding for summer • (applications due January 2010)